Towards Understanding Adversarial Transferability in Federated Learning

Yijiang Li, Ying Gao, Haohan Wang

We investigate a specific security risk in FL: a group of malicious clients has impacted the model during training by disguising their identities and acting as benign clients but l…